Abstract
Antidepressant drugs are held to increase the risk of manic ‘switching’ in depressed patients (both following their initiation and discontinuation) as well as increase the chance of rapid cycling. The possibility of antidepressantinduced switching appears greatest in patients with bipolar depression, and probably in those with bipolar I disorder. A number of papers have addressed the management of bipolar depression over the last decade [e.g. 1–4], all considering the issue of switching in some detail, and contemplating the issue addressed here: whether differing antidepressant classes effect differential risks to switching. Concerns about switching make many clinicians reluctant to prescribe an antidepressant when a depressed patient has a history of bipolar disorder. Moller and Grunze [5] have noted how such concerns have led some US and Canadian guidelines to recommend against antidepressants in managing bipolar depression and to favour mood stabilizers, a recommendation that the authors viewed as excessively restrictive and unbalanced.
This paper reviews the general topic of switching and considers whether avoidance of certain antidepressants and the pre-emptive strategy of prescribing a mood stabilizer prior to commencing certain antidepressants is necessary to the management of bipolar depression. We review relevant reports and reviews.
An historical overview of antidepressant-induced ‘switching’ and rapid cycling
Reports of the capacity of antidepressants to induce manic switching emerged in the 1960s but were initially rejected [6]. Thus, while Angst [7] reported that the switch rate from depression into mania for inpatients of a Zurich psychiatric hospital rose from 3.0% in the preantidepressant period (1920–1959) to 9.2% following the introduction of antidepressants in 1959, he concluded that there was ‘no evidence of a treatment-induced switch’. Lewis and Winokur [8] undertook a chart review in the early 1980s and held that tricyclic antidepressant (TCA)-induced switching ‘probably represents random manifestations of bipolar illness’. Thus, as switching is clearly part of the natural history and definition of bipolar disorder, the phenomenon was unable to be demonstrated (or accepted by clinicians) until firm evidence indicated that rates of induction of mania were higher in depressed subjects taking antidepressants compared with those not so exposed.
Representative studies over the next two decades suggested quite high switch rates with the then used broader-spectrum antidepressants, with one report [9] giving rates of 21% and 25% with the monoamine oxidase inhibitor (MAOI) tranylcypromine and the TCA imipramine, respectively. Bipolar I (as against bipolar II) depressed patients were held to be at particular risk of switching [10]. Reviews [e.g. 11]. variably suggest that TCA-induced switching may or may not be prevented by use of a concurrent mood stabilizer, but there does appear to be empirical evidence [12, 13] supporting lithium as reducing the risk.
Subsequent to multiple reports suggesting that the TCAs and MAOIs cause switching, case reports have described manic switching in conjunction with all of the new antidepressants, including bupropion [14], citalopram [15], fluoxetine [16], fluvoxamine [17], mirtazapine [18], moclobemide [19], paroxetine [20], sertraline [21], trazadone [22] and venlafaxine [23]. Further, we have reported [24] that augmentation of antidepressants with the atypical antipsychotic olanzapine may induce manic switching. Such findings encouraged a clinical view (or myth) that any effective antidepressant should possess the capacity to induce switching.
A long-standing and unresolved question is whether switching occurring in those without any previous manic episode requires a diathesis to bipolar disorder or is independent of any such predisposition. The latter possibility is indirectly supported by the suggestion [e.g. 25] that antidepressant-induced mania is not ‘true mania’, in being less severe and of shorter duration than spontaneous mania. Another issue requiring resolution is whether antidepressants induce manic episodes in some individuals who would otherwise have unipolar histories.
In terms of ‘discontinuation switching’, quite differing antidepressant drugs have been incriminated, including paroxetine [26], fluvoxamine [27], venlafaxine [28] and the tricyclics or TCAs [28]. It remains unclear as to whether any such switching is a direct cause of the antidepressant being withdrawn, is induced by removal of any prophylactic effect of the antidepressant, or again reflects the natural history of bipolar disorder.
In terms of the ‘induction of rapid cycling’, longitudinal studies of bipolar patients [e.g. 29] indicate that both intense and protracted use of antidepressant drugs has been implicated as causing transformation to a course of rapid cycling, with antidepressant-induced mania being suggested [30] as a risk marker for future rapid cycling. Wehr and Goodwin [31] suggested that TCA-induced rapid cycling reflected the TCA ‘accelerating’ the ‘natural, cyclic course of the illness in all its phases’.
We now focus on the first issue and, in particular, pursue whether antidepressant-induced manic switching is likely to be caused by the ‘newer’ antidepressant drugs. As the data in regard to switching appear to differ in relation to unipolar and bipolar depression, each will be considered separately.
Unipolar depression
A review published by Peet in 1994 [32] produced seemingly unequivocal data in suggesting differential switch rates for TCAs and SSRIs in unipolar depressed patients. Peet aggregated clinical trial data for four SSRIs (i.e. fluoxetine, fluvoxamine, paroxetine and sertraline) together with comparator TCA and placebo data. Manic switch rates were 0.52%, 0.72% and 0.21% for those receiving TCA, SSRI and placebo, respectively. Thus, these data suggest that unipolar patients may be slightly more likely to switch if receiving a TCA or an SSRI (in comparison to receiving a placebo), a result achieving formal significance partly because of the large database (involving more than 16 000 patients). However, the switch rate was less than 1% across each of the three groups. Any such low incidence phenomenon risks misinterpretation. Statistical significance may be achieved by a large sample size, while statistical non-significance risks (for a low base-rate phenomenon) illustrating the aphorism that ‘absence of proof is not proof of absence’. Peet wisely stepped around this issue by stating ‘In unipolar patients, differences between the two classes of antidepressant and placebo are not large enough to have any effect upon clinical practice’.
Thus, can we conclude that the SSRIs (as an example of a new antidepressant class) do not cause switching in those with unipolar depression? Possibly yes, if we restrict consideration to grouped data from such controlled studies, and in relation to low or ‘standard’ dosages. However, there are a significant number of case reports describing switching in conjunction with high SSRI doses. Representative reports include switching following sertraline at 300 mg/day, paroxetine at 80 mg/ day [33], fluoxetine at 80 mg/day [34] and of more than 100 mg/day [35, 36], and with mania settling on lowering of the dose.
Bipolar depression
There have been many estimates of the switch rate in untreated bipolar patients [e.g. 8], but they are of little utility as sample characteristics (particularly illness components) ensure that any estimate is idiosyncratic to that sample. Thus, it is necessary to examine comparative switch rates in bipolar patients exposed to antidepressant medication and to placebo.
In the Peet [32] review of clinical trial data for SSRIs, comparator TCAs and placebo, the rates for switching were 3.7%, 11.2% and 4.2%, respectively. Compared with the data for unipolar patients, the bipolar set therefore showed higher switch rates than the unipolar set, whether receiving an SSRI (3.7% vs 0.52%), TCA (11.2% vs 0.72%) or placebo (4.2% vs 0.21%), suggesting a distinct natural history contribution to switching in those with bipolar depression. The data further suggested that bipolar patients were more likely to switch if receiving a TCA but no more likely if receiving an SSRI (arguing for a drug class specificity effect). Such data, however, remain indicative. First, as the drug trials were undertaken over brief intervals; although ‘switching’ is commonly viewed as a more immediate rather than a delayed phenomenon. Second, as Peet did not distinguish between bipolar I and II categories in his analyses, and when there are some reports suggesting that switching is more likely in those with bipolar I disorder compared with bipolar II patients [9, 10, 37].
Differential switch rates across differing antidepressant classes have certainly been demonstrated for those with bipolar I depression. For example, Calabrese and colleagues [38] (principally examining data for moclobemide, imipramine and paroxetine) confirmed that TCAs presented the highest risk to switching. This leaves the question of differential switch rates in bipolar II depression open, reflecting the absence of any substantive data set for this diagnostic group. As reported in this journal [39], a small case series allowed the suggestion that a percentage of bipolar II patients commenced on an SSRI or venlafaxine for a depressive episode actually reported manic swings as being less common or less persistent, suggesting that such drugs might actually have some mood stabilizing potential (rather than inducing switching or cycling) in bipolar II patients.
The indicative data (for bipolar patients as a whole) have resulted in expert consensus guidelines [40] recommending the use of two of the new antidepressant classes (i.e. bupropion and SSRIs) for treating bipolar depression.
Switching in association with the ‘broader action’ antidepressants
The mechanism underlying ‘natural switching’ remains unestablished, but Goodwin and Jamison [41] imputed increased noradrenergic and dopaminergic functioning, and provided some evidence for norepinephrine-enhancing drugs producing manic switches. If valid, then drugs with a noradrenergic action might be expected to be associated with a higher rate of switching than ones with preferential serotonergic actions.
As yet, no detailed analyses have been undertaken of the newer antidepressant classes that have noradrenergic actions, so that we can only rely on data from clinical trials which, by design, usually exclude bipolar subjects. In regard to the SNRI venlafaxine, Wyeth (Australia) reported [Fox J: personal correspondence] that those with bipolar disorder were excluded from pre-marketing trials, where ‘mania or hypomania occurred in 0.5% of unipolar depressed patients treated with Efexor and in 0.3% of patients treated with Efexor XR’. The company noted studies of venlafaxine for bipolar depressed patients. In a single-blind study [42], 60 outpatients with bipolar depression received venlafaxine or paroxetine, and with 4 patients (13%) and 1 patient (3%), respectively, having a manic or hypomanic switch. In a second double-blind study, venlafaxine was assessed in 17 bipolar II and 31 unipolar patients [43] with no episode of manic switch or mood cycling being reported. In an unpublished study, Gastpar and colleagues assessed the safety and efficacy of venlafaxine in 176 patients with bipolar disorder, with a switch rate of 3.3% over the 12 weeks. In regard to the NaSSA Mirtazapine, Organon (Australia) reported [personal correspondence] that a ‘hypomanic or manic switch occurred in 0.25% of all mirtazapinetreated patients during the worldwide clinical trial programme’ [44].
Conclusion
The data reviewed here suggest first that antidepressants of all types are either unlikely to induce switching in unipolar patients or provide a low risk at best. For bipolar patients, the risk of switching appears distinctly increased for broad spectrum TCAs and MAOIs, but does not appear increased in association with the SSRIs. Trial data also suggest that the dual action drugs such as mirtazapine and venlafaxine are unlikely to present an increased risk, but it would be unwise to rely only on clinical trial data.
Many current recommendations are of concern as they do not appear to reflect the knowledge base reviewed here. In essence, excessive concerns about antidepressant-induced switching appear to have driven extremely conservative recommendations. Thus, for the acute treatment of bipolar depression, Nolen and Bloemkolk [45] suggested that clinicians should first prescribe a mood stabilizer and, if no response, add an antidepressant only after 4–6 weeks. US Expert Consensus Guidelines [40] recommend a mood stabilizer and an antidepressant. Risks of delay and redundancy in treating depression are evident in such recommendations.
We suggest then that clinicians should be less concerned about the risk of switching if managing a bipolar depressed patient with a narrow action antidepressant such as an SSRI. However, while the SSRIs appear less likely to so cause harm, they may nevertheless be less effective in such circumstances. Non-psychotic bipolar depression is most commonly melancholic in type [46], and the efficacy of the narrow action antidepressants compared to broader antidepressants to treat melancholic depression in older patients appears less powerful [47]. Reconciling the common need to use a broader action antidepressant in such circumstances then needs to be balanced against the greater chance of inducing switching. Such nuances will continue to require considered clinician judgement as they do not appear resolved or adequately addressed within current guidelines.
Footnotes
Acknowledgements
Thanks to Kerrie Eyers for editorial assistance, and the NHMRC (Program Grant 222708) and the NSW Centre of Mental Health for Institute and Infrastructure Funding.